Natural Raises $30M to Build Payment Rails for Agents That Move at Computer Speed

Natural, a two-year-old startup building payment infrastructure for AI agents, has raised a $30 million Series A led by Forerunner, with founder Kirsten Green personally leading the round. Combined with prior funding, that brings the company's total raised to $40 million — a meaningful bet on a thesis that's still mostly unproven at scale: that autonomous agents need financial rails built for them from the ground up, not human-authorization flows retrofitted to accept agent traffic.

The problem Natural is targeting is structural. Existing payment infrastructure assumes a human is present to authorize a transaction — enter a card number, approve a push notification, confirm a purchase. That model works fine when an agent's job ends at "here are three vendor options, pick one." It breaks down the moment an agent is expected to identify a vendor, negotiate or compare pricing, and execute the purchase end to end, because someone still has to sit in the authorization loop, capping the process at human speed. Natural's pitch is an orchestration layer that lets agents move money, collect it, and transact with humans or other agents without that human-speed bottleneck.

Co-founder and CEO Kahlil Lalji, who started the company in 2025 with Eric Wang (previously of Ivella) and Walt Leung (a former Nextdoor engineering manager), is positioning Natural as a direct challenger to Stripe, which is running its own push into agent-native payments. Lalji's framing is that Natural's early architectural choices — built agent-first rather than retrofitted — give it "a good shot" against larger incumbents with deeper distribution. The company is also weighing USD-backed stablecoin rails alongside traditional bank-payment support, a hedging move that puts it in more direct competition with Skyfire Systems, another agent-payments startup working the stablecoin angle.

The bigger signal here isn't Natural specifically — it's the volume of capital now chasing the same wedge from different directions. Stripe is defending its incumbency by building agent support into existing rails. x402 and similar HTTP-native micropayment standards are attacking the problem from the API-call layer, letting agents pay per request instead of per checkout. Natural and Skyfire are attacking it from the orchestration layer, treating agent-to-agent and agent-to-vendor commerce as a distinct product category rather than a feature bolted onto human checkout. None of these approaches has established itself as the default yet, and they're not strictly competing on the same surface — an agent could plausibly pay for an API call via x402 and settle a vendor invoice via Natural in the same workflow. Whoever ends up owning the most transaction volume will likely be decided less by architecture and more by which rail gets embedded into the agent frameworks developers actually ship with.